People pick out right angles, parallel sides and symmetry without being taught, and that intuition is treated as a foundation of elementary Euclidean geometry in us. Whether it is ours alone has been hard to ask, because most animal tests need training on each new case. Schmidbauer and colleagues trained on a general odd-one-out task with arbitrary shapes, then handed them they had never seen. What did the birds do?
They applied it immediately, and they were better on the regular shapes. The crows picked out the quadrilateral with different geometric properties straight away, with no training on quadrilaterals at all. Performance was better when the shapes carried right angles, parallel lines or symmetry than when they were more irregular, and that advantage did not have to be learned. Geometric intuition is not unique to us.
Show a person five shapes and one odd one, and they find it without instruction. Right angles, parallel lines and symmetry carry a pull that feels like perception rather than reasoning, and that pull is treated as an elementary piece of mathematical intuition in humans.
Asking an animal is harder than it sounds. Train a bird on squares and it may simply learn squares, which tells you nothing about geometry. The question needs a task the animal already understands, into which new shapes can be dropped.
Schmidbauer and colleagues trained carrion crows to detect a visually distinct among six arbitrary shapes, which is a general concept rather than a specific rule. Then they handed the birds sets of quadrilaterals, where the odd one out differed from the rest only in its geometric properties. The crows applied the concept immediately.
The second finding is the sharper one. The crows showed a effect, doing better with shapes that had right angles, parallel lines or symmetry than with more irregular ones, and that performance advantage did not require learning. Geometric intuitions, on this evidence, are not specific to humans but rooted somewhere deeper in evolution.
wat. Nobody taught the crows what a right angle is, and right angles were the ones they found easiest.
wat wat
It pushes a basic mathematical intuition back from human culture into biological evolution, in a lineage that split from ours a very long way back.
How do we know?
Schmidbauer, Hahn and Nieder trained carrion crows to spot a visually distinct intruder among six arbitrary shapes, then tested transfer to sets of quadrilaterals differing in geometric properties.
What is still unknown
One species and one family of shapes. Detecting that a quadrilateral differs geometrically is not the same as reasoning about geometry, and the paper does not claim it is.
Learn something?
Where to next
Can a bird decide in advance how many times to call?crows
Can an animal build a tool out of pieces that are each useless?crows
Does a crow read a face the way we do?crows